Watch a tense scene from any film with the sound off and it stops being tense. Watch it with the picture off and most of the tension is still there. Yet very little of that sound was recorded on the day: the microphone on set was pointed at the actors' mouths, and everything else — the footsteps, the wind, the door, the music, even the silence — was built afterwards, layer by layer, and then balanced so that the audience hears exactly what the story needs at every moment. That building and balancing is sound design and mixing.
The layers of a soundtrack
| Layer | What it is | Why it is there |
|---|---|---|
| Dialogue | The voices, from the set microphones or re-recorded in a studio | Carries the story. Everything else is balanced around it |
| Room tone | The quiet sound of the actual location with nobody speaking | Fills gaps and joins so silences match and edits are invisible |
| Ambience (atmos) | Background sound of the wider world: traffic, birds, a cafeteria | Tells the audience where they are, even off-screen |
| Foley | Everyday sounds performed in sync with the picture: footsteps, clothing, props | Makes bodies and objects feel physically present |
| Hard effects (SFX) | Specific, often library sounds: a car door, a phone buzz, an explosion | Marks events the camera shows or implies |
| Music | Score or licensed tracks | Sets emotion and pace, and tells the viewer how to feel |
Foley is named after Jack Foley, a sound effects artist at Universal Studios who developed the technique of performing sounds live to a projected picture. Foley artists walk in shoes on trays of gravel, rustle fabric for every movement (a cloth pass), and handle props in time with the actors. It exists because the set microphone captures dialogue clearly but footsteps and clothing faintly and inconsistently, and because a sound that is recorded separately can be controlled separately in the mix. It also allows an M&E track (music and effects, without dialogue), which is what lets a film be dubbed into another language with all its other sound intact.
Mixing: dialogue first, everything else around it
- Work in stems. Group tracks into dialogue, music and effects buses so each family can be adjusted with one fader — and delivered separately if a client asks.
- Set the dialogue first so it peaks consistently, then bring every other layer up underneath it. A mix that starts with the music almost always ends with a buried voice.
- Clean the voice with EQ. A high-pass filter around 80 Hz removes rumble from traffic, air handling and microphone handling, none of which a speaking voice needs.
- Compress gently. A compressor narrows the gap between a speaker's loud and quiet words so the quiet ones stay audible over the music.
- Keep dialogue in the centre. Pan ambience and effects to match where they are on screen, but the voice stays centred so it survives a phone speaker or a mono mix.
- Fade every edit. A few frames of fade on each audio cut prevents the clicks that appear when a waveform is chopped mid-cycle.
Sample rate and bit depth
Digital audio measures the sound wave many times a second and stores each measurement as a number. The sample rate is how many measurements are taken per second; the highest frequency it can capture is half the sample rate (the Nyquist limit), so 48 kHz captures up to 24 kHz — beyond the roughly 20 kHz limit of human hearing. 48 kHz is the standard for video; 44.1 kHz is the CD and music standard, and mixing the two in one project forces a conversion. The bit depth is how precisely each sample is stored. Each bit adds about 6 dB of dynamic range, so 16-bit gives about 96 dB and 24-bit about 144 dB. Recording at 24-bit is why you can afford to leave generous headroom: the quiet parts still sit far above the noise floor.
Delivery loudness and music licensing
| Destination | Integrated loudness target | Notes |
|---|---|---|
| Canadian and US television | About −24 LKFS | The ATSC A/85 practice, which the CRTC adopted so commercials are no louder than programs |
| European broadcast (EBU R 128) | −23 LUFS | True peaks no higher than −1 dBTP |
| Streaming video and music | Roughly −14 LUFS | Platforms normalize: louder uploads are simply turned down |
| Podcasts | Around −16 LUFS | A common recommendation for spoken word |
- A recorded song carries two copyrights: the composition (melody and lyrics, usually controlled by a publisher) and the master recording (usually controlled by a label). Putting it in a video needs a sync licence for the first and a master use licence for the second.
- Owning or streaming a song is not permission. A purchase or subscription is a licence to listen privately, not to synchronize the music with your pictures and publish them.
- Royalty-free is not free. It means a one-time fee instead of ongoing payments, and the licence still limits where, for how long and in what kind of project it can be used.
- Creative Commons tracks are usable within their conditions: credit the artist (BY), and check for NC (no commercial use) and ND (no derivatives). Creative Commons treats putting music in sync with video as making an adaptation, so ND tracks are generally off-limits for video soundtracks.
- An old composition can have a new recording. Beethoven's music is in the public domain, but a 2015 orchestra's performance of it is protected. Use a public-domain recording, or record it yourself.
- Keep the licence. Automated systems such as YouTube's Content ID match audio against registered recordings; the licence document is what you show when a claim lands.